Predicting pKa of Amines for CO2 Capture: Computer versus Pencil

Article pubs.acs.org/IECR

Predicting pKa of Amines for CO2 Capture: Computer versus Penciland-Paper Kazi Z. Sumon,† Amr Henni,† and Allan L. L. East*,‡ †

Faculty of Engineering and Applied Sciences, and ‡Department of Chemistry and Biochemistry, University of Regina, Regina, Saskatchewan S4S 0A2, Canada S Supporting Information *

ABSTRACT: For pKa prediction of aliphatic amines (amines relevant to industrial CO2 capture), the performance of quantumchemistry continuum-solvation methods is contrasted with the 1981 pencil-and-paper group-additivity method of Perrin, Dempsey, and Serjeant (PDS). We have optimized a quantum-chemistry continuum-plus-correction strategy, as well as updated the parameter values in the PDS method. The PDS method outperformed the continuum-based method: root-mean-square errors for a sample of 32 amines are 0.28 for the continuum-based method, 0.33 for the original PDS method, and 0.18 for the updated PDS method. Considering also that there is ambiguity in choice of cavity radii and molecular conformer for continuumbased methods, we recommend the pencil-and-paper PDS method over such methods.

1. INTRODUCTION There is need for a simple procedure that can predict pKa to within ±0.3 for a variety of amines important in CO2 capture.1,2 The pKa of a compound has been estimated via quantum chemistry calculations using fairly routine continuum-solvation approximations, but the accuracy of these models is often unsatisfactory:3,4 since 2005, most such procedures for studies of 10 or more amines have seen pKa errors of ±1 or greater,5−7 and the addition of correction terms, either constant7,8 or proportional to the pKa magnitude,5,6,9−13 has reduced errors but not to ±0.3 accuracy. At the moment, the best of these methods have pushed accuracy to ±0.4 for selected amines8,12 (±0.2 for the subclass of benzimidazoles5). Some researchers have bypassed continuum-solvation methods in favor of linear correlation formulas to other computable properties of the amine,14,15 but to date have not improved upon ±0.4 accuracy. We published in 2009 a general continuum-plus-correction procedure for pKa computation, which produced rms errors of 0.68 for a test set of 17 amines.7 It employed Gaussian 0316 self-consistent reaction-field (SCRF) computations in a semicontinuum manner (“Model II”), in which an explicit water molecule is inside the cavity with the solute molecule: the H2O is arranged to H-bond to the lone pair on N in the neutral amine, but to receive an H-bond from the protonated amine. The procedure, now labeled the KHE method, incorporated many terms in the calculation, including a statistical entropy effect for multiple conformations, in an attempt to get as much accuracy as one could out of the dielectric-continuum approximation. We have now simplified the procedure drastically, for both Gaussian 03 (G03)16 and Gaussian 09 (G09)17 users, and pushed its accuracy down to a root-meansquared error of 0.28 for a training set of 32 CO2-relevant amines (acyclic amines including alkanolamines, and substituted piperazines and morpholines). There remain problems, particularly with other classes of ringed amines. The new procedure is labeled SHE to distinguish it from our earlier one. © XXXX American Chemical Society

However, there is a computer-free group-additivity method for pKa prediction, published in a book by Perrin, Dempsey, and Serjeant (PDS) in 1981,18 that was known to provide accuracy to a few tenths for amines. Their ΔpKa method was based on the knowledge that use of Taft parameters for organic substituents in an additive way can account for trends in equilibrium constants and free energies of reaction within classes of compounds.19 This Article compares pencil-andpaper PDS results against those of the computer-based SHE method.

2. METHODS 2.1. SHE Method. The pKa of a base B is a scaled version of ΔrG(aq), the free energy change of the acid-dissociation reaction BH+(aq) → B(aq) + H+(aq). With energies in kcal mol−1, the relation at T = 298.15 K is pK a =Δr G(aq)/RT ln 10 = Δr G(aq)/1.3643 = [G(aq)(H+) + G(aq)(B) − G(aq)(BH+)]/1.3643 (1)

These Gaq quantities can be approximated with dielectriccontinuum models, by breaking each Gaq down into a sum of three components (Eel + ΔGnonel + Enuc): Briefly, Eel = ⟨Ψ( f)|H + V/2|Ψ(f)⟩ is the internal energy of the solute in solution (including its electrostatic interaction with the solvent continuum), ΔGnonel is the correction due to nonelectrostatic contributions to solvation (cavitation, dispersion, repulsion, solvent-structuring effects), and Enuc is the nuclear-motion energy of the solute in solution. Such computations are Received: April 20, 2012 Revised: August 12, 2012 Accepted: August 22, 2012

A

dx.doi.org/10.1021/ie301033p | Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Industrial & Engineering Chemistry Research

Article

problematic for G(aq)(H+) due to the strong covalent nature of H+ in solution (H3O+, H5O2+, etc.), and hence G(aq)(H+), a constant with respect to identity of base, should be estimated differently. For the other two free energies, we consider the half reaction free energy ΔhG(aq)(BH+ → B) and its components (ΔhEel + ΔhΔGnonel + ΔhEnuc), leading to

Model I (no explicit water) gas-phase frequency values, which were consistently −9.4 ± 0.5 kcal mol−1 (Table 6 of ref 7). Closer inspection revealed that this amount is dominated by the zero-point vibrational energy (ZPVE) term: deprotonation generally resulted in the loss of a 3400 cm−1 NH stretch and two 1600 cm−1 HNC bend modes, and no net frequency shifts of other modes. Friesner and co-workers had already commented on how constant this term likely is for a particular class of compounds.22 ΔhΔGnonel = 0. We remind Gaussian users that three of the four nonelectrostatic effects (cavitation, dispersion, repulsion) are computed by default in G03 but must be requested in G09, and in G03 the values appear only in the middle of the logfile. Their effects are generally small. The sum of these three effects for the half-reaction, ΔhΔGnonel, was computed (G03) to be small and nearly constant: 0.2 and 0.6 kcal mol−1 in Model I and Model II calculations, respectively (Tables 6 and 7 of ref 7). These small shifts tend to be swamped by remaining errors, and hence not of sufficient benefit for inclusion in a simple procedure. Empirical Corrections. To make the errors uniformly small for multiple classes of compounds, class-dependent empirical corrections C are needed (see section 3.1). Also, in general, optimal values for C would depend on the choices made for computing ΔhEel: level of theory, cavity radii, and molecular conformer. Attempts were made to find theoretical justification for a choice of cavity radii and molecular conformer, but these failed (section 3). Given the choices we have made for ΔhEel, values for C deemed optimal for a training set of 32 amines (Table 5, section 4) are −0.7 for acyclic amines and −1.7 for cyclic amines. In summary, the SHE procedure is to compute

pK a = [G(aq)(H+) + Δh Eel + Δh ΔGnon‐el + Δh Enuc] /1.3643

(2)

All four terms in eq 2 have been simplified in going from KHE to SHE, as follows. G(aq)(H+) = −270.3 kcal mol−1. The KHE method used this G(aq)(H+) term as a fitting parameter, because it is difficult to predict ab initio. We now adopt a proper value, the sum of G(g)(H+, 1 atm) = −6.3 kcal mol−1, computed via a Sackur− Tetrode equation, with ΔsolvG(H+, 1 atm gas → 1 M aqueous) = −264.0 kcal mol−1, derived by Tissandier et al.20 from experimental data. Using the Ben-Naim convention21 for ΔsolvG, one obtains the same value, summing G(g)(H+, 1 M) = −4.4 kcal mol−1 with ΔsolvG(H+, 1 M gas → 1 M aqueous) = −265.9 kcal mol−1. ΔhEel Now Employs MP2/6-31G(d). The KHE method used MP2/6-311++G(d,p), maximally trans conformers for acyclic amines, and UA0 radii for solute cavities. However, further basis set testing by us revealed that the removal of diffuse functions actually reduces errors for acyclic bases in both Model I and Model II calculations (without or with an explicit H2O inside the cavity, respectively); Model II results are shown in Table 1. We also decided to reduce the basis set from triple- to Table 1. pKa Errors from Uncorrected SHE Procedure: Basis Set Dependencea amine

6-311++G(d,p)

6-311G(d,p)

6-31G(d)

NH3 MeNH2 Me2NH Me3N piperazine

−4.24 −2.81 −2.14 −1.88 −0.77

−0.06 0.78 1.14 1.36 2.48

−0.49 0.52 0.67 0.48 2.37

pK a = (1/1.3643)[ − 270.3 + Eel(B·H 2O) − Eel(BH+· OH 2) − 9.4] + C

(3)

where the Eel(B·H2O) and Eel(BH+·OH2) are MP2 energies, converted to kcal mol−1 (×627.50955), from the bottom of Gaussian logfiles (“MP2=”) of SCRF=(PCM,G03Defaults) MP2/6-31G(d) geometry optimizations of maximally trans conformers. The C value is −1.7 for cyclic amines and −0.7 for acyclic amines. 2.2. PDS Method. The empirical pencil-and-paper Perrin− Dempsey−Serjeant method18 uses only a table (Table 2 here) of pKa “base” values and ΔpKa additive functional-group corrections, thus offering much faster predictions. The original publication also offers some temperature correction formulas for pKa. Because the parameter values in the PDS scheme were chosen before much alkanolamine data had been studied, we provide an updated set of these values (“new PDS”), obtained from a least-squares fit to experimental pKa values of the same 33-amine training set used for SHE. Please note that we have only updated the parameter values relevant for CO2-capture amines; the original method offers terms for other amines, and carboxylic acids as well.18 Table 2 lists the old and new parameter values. Of the changes, note that (i) the base values for primary and tertiary amines are now equal, and (ii) the correction for cyclic amines is now zero. Our SHE results actually reveal a ring effect of −1.0 in the experimental pKa values (section 3.1); in the PDS method, the effect is incorporated by counting a β group twice if it occurs in a ring.

a

SHE method as described in this work, but omitting empirical corrections and allowing for basis set substitution in the SCRF MP2 geometry optimizations. Piperazine calculations employed conformer choice of ref 7. Experimental pKa values are 9.25, 10.66, 10.73, 9.80, and 9.73 for the five bases (CRC Handbook23).

double-ζ, possibly sacrificing some accuracy but gaining in speed for extension to larger amines. The UA0 radii are default in G03; in G09 they are requested by adding G03Defaults to the SCRF keyword. ΔhEnuc = −9.4 kcal mol−1. We dramatically simplify this nuclear-motion energy term. Enuc for a solute has an intrinsic term, Enuc,int, and a statistical entropy term −TSstat due to multiple conformations and rotational symmetry.7 In the KHE method, Sstat was derived for each solute (B and BH+) from symmetry considerations and conformer counting, while Enuc,int was computed from vibrational frequencies of Model II gasphase complexes (i.e., with solute·H2O complexes). We now neglect TSstat effects altogether, as the effects are less than 1 kcal mol−1 (Table 3 of ref 7), and are a hassle to incorporate for the nonspecialist. We also now replace ΔhEnuc,int with the constant −9.4 kcal mol−1 to eliminate the need for gas-phase opt+freq calculations. The value −9.4 was chosen to mimic previous B



Article

Table 2. Terms in the Perrin−Dempsey−Serjeant Scheme for pKa Predictiona term base value

ΔpKa shifts

primary amine NH2R secondary amine NHR2 tertiary amine NR3 each CH3 on N each β OR each β NHR or NR2 each β OH each β NH2 each γ group each δ group ring effect if two equivalent N sites

original18

updated

10.77 11.15 10.50 −0.2 −1.2 −0.9 −1.1 −0.8 +0.4Δβ +0.4Δγ +0.2 +0.3

10.6 11.1 10.6 −0.2 −1.4 −1.0 −1.0 −0.9 +0.4Δβ +0.4Δγ 0 +0.3

a

Terms for aliphatic N- and O-containing amines; for other groups and molecules, see the original method. The β effect is added twice for ringed compounds such as morpholine and piperazine. A γ effect is considered 40% the magnitude of a β effect; for example, for monopropanolamine, pKa = 10.6 + 0.4 × (−1.0) = 10.2 using updated parameters.

Figure 2. Cavity volumes of B·HOH complexes. See Figure 1 for legend.

smallest volumes (23 and 19 Å3, respectively; the other procedures use volumes ≥26 and 21 Å3, respectively).28 Regardless of choice of algorithm and radii, there are classdependent errors evident in Figure 1, particularly the unique errors for cyclic compounds (morpholine and the piperazines in our case) that have been seen before.7,29 Friesner and coworkers hypothesized29 that the error shift for cyclic compounds is due to neglecting the fourth type of nonelectrostatic effect: a solvent-structuring effect. Such solventstructuring may be clathrate-like;30 solid clathrate structures are known to be stable around cyclic ethers, for example.31 This necessitated the use of separate values for the empirical shift C for cyclic versus acyclic amines (section 2). 3.2. Choice of Conformer. An additional unresolved issue is the choice of molecular conformer. This commonly ignored problem is important because pKa results for alkanolamines are heavily dependent upon conformer choice (Table 3).32 In the examples in the table, conformer effects are ∼0.5 for ethyl group rotation, but they are 2−4 for alkanolamine internal rotation. The large effect for alkanolamines is due to gauche XCCY conformations, which place the polar X and Y groups in close proximity; this preferentially stabilizes the BH+ cation in a continuum-solvation calculation, and shifts up the pK a prediction roughly 2 units per polar gauche interaction. Note that these large pKa variations with internal rotation of an XCCY unit are as large as the variations in switching default cavity radii in Figure 1. In fact, if G09 default radii (UFFx1.1) are used, Table 3 reveals that one gets the best agreement with experiment if gauche conformers are used instead of trans ones! We chose the trans + UA0 combination over the gauche + UFFx1.1 combination mainly because UFF radii show less uniform errors than UA0 radii in Figure 1, which translates into worse rms errors as one tests more amines. It would have been better, of course, to choose the conformer based on which conformer dominates in aqueous solutions. Unfortunately, theory has been unable to predict this to date: (i) the rule of thumb that one should use the lowest-energy conformer in a conformer search cannot be employed, because although gasphase and dielectric-continuum optimizations both predict the gauche forms of both B and BH+ to be lower in energy than the trans forms, this prediction neglects the effects of explicit hydrogen bonding with solvating water molecules;34 (ii) two nanosecond-scale classical dynamics simulations of aqueous monoethanolamine (MEA), which purported to demonstrate

3. CONTINUUM-SOLVATION ISSUES 3.1. Choice of Radii. The accuracy of continuum-solvation computational methods is, unfortunately, very sensitive to the atomic radii used to define the solute cavity,24,25 and we noted that the default radii choice changed (from unscaled “UA0” to “UFF” scaled by 1.1) when Gaussian upgraded its software from G03 to G09. In Figure 1, the change shifts pKa predictions

Figure 1. SHE results without empirical corrections (on conformers of ref 7), showing dramatic effects of cavity radii. ◆: experiment. ■: G03 (UA0). □: G09 radii=UA0(x1.1). ▲: G03 radii=UFF. △: G09 (UFFx1.1). Parentheses denote default effects.

down by an alarming amount: 3−4 units. (There was also a technical change in cavity and solute−solvent−surface construction in going from G03 to G09, which may have had a small effect upon predicted pKa.26,27) In Figure 1, good estimates for trans conformers of all acyclic amines (NH3 to DEA) were achieved with the unscaled UA0 radii, and hence this choice was incorporated into the SHE procedure. Although the UA0 choice provides the highest pKa estimates of the choices in Figure 1, it provides only moderate total molecular volumes (Figure 2); the key seems to be in the volume given to the reactive site, particularly primary and secondary N atoms, where the UA0 choice results in the C



Article

observed NCCO dihedral angles were reported in a third such study.37,38 To our knowledge, the best relevant experimental study of this issue is a 1975 Raman spectrum study,39 which demonstrated that both conformers of neutral MEA are present in solution; this calls into question the habit of choosing only one for quantum chemistry computation.

Table 3. pKa Results from Uncorrected SHE Procedure: Conformer Dependence aminea

conformerb

pKa G03c

pKa G09c

AMP 9.68

G T

11.3 9.8

7.9 6.1

DEA 8.88

GtgG GttG TtgG TgtG TttG TggT TgtT TttT

12.4 11.9 11.2 10.8 10.8 9.6 9.5 9.5

9.3 8.9 7.6 7.4 7.3 5.8 5.7 5.7

DIPA 8.84

GttG TttG TttT

12.5 10.5 8.6

9.1 7.3 5.3

MEA 9.5

G T

11.7 10.0

8.2 6.3

Et2NH 10.84

gg tg tt

12.2 12.1 11.9

9.4 9.1 8.9

PPZ 9.73

GG TT ax/eq TT ax/ax TT eq/eq TT eq/ax

12.5 12.1 12.1 11.7 11.6

9.2 8.5 8.5 8.0 7.9

0.9

1.1

rms error

4. RESULTS If one accepts the limitations inherent in a simple oneconformer procedure with a traditional continuum-pluscorrection methodology, the SHE method constitutes an improvement within its class, as supported by a direct comparison of SHE results to those of other continuumsolvation procedures (Table 4). However, SHE is outperformed by the “New PDS” method. First, Table 5 presents PDS and SHE results for the 32-amine training set. Both SHE and the “New PDS” methods were trained on this set and give rms errors of 0.28 and 0.18, respectively. Second, Table 6 presents further comparisons of SHE and New PDS results, for 11 amines outside of the training set, and while the New PDS method is still performing very well, the SHE method shows further weaknesses, particularly with piperidine and 1-methylpyrrolidine. Looking more closely, the SHE0 calculation (without empirical shifts) reveals an error that is dependent on the class of ringed compound: +1.9 for piperazines, +1.4 for morpholines, +1.2 for piperidines, and +0.8 for pyrrolidines. It seems that the entropy change of the surrounding water upon protonation of the amine (the only term not in the calculation) is important in an ab initio calculation of pKa for ringed compounds. Rather than introducing more empirical shifts onto a continuum-solvation electronic structure procedure, one should turn to a simpler procedure that has only empirical shifts and abandons the electronic structure calculation altogether, that is, the empirical PDS methods.

a

Experimental pKa values are listed: from Hamborg and Versteeg33 for AMP, DEA, and DIPA, and from CRC Handbook23 for MEA, Et2NH, and piperazine. bGauche and trans descriptors for NCCX dihedrals (G,T) and CNCC dihedrals (g,t) along the main atom chain. The HOCC and PNCC (P = lone pair) dihedrals are not specified in this notation; for T cases, we took them to be trans, but for G cases they were chosen to arbitrarily maximize NH···O hydrogen-bond interaction. For PPZ, the axial and equatorial descriptors are for the NH bonds of protonating and spectator N atoms, respectively. cBold values represent the best single-conformer predictions; the difference between G03 and G09 calculations is the default cavity radii used (UA0 and UFFx1.1, respectively).

5. CONCLUSION For predicting aqueous pKa values of CO2-relevant amines, we have pushed the continuum-plus-correction method about as far as we can push it, producing a method (SHE) whose pKa predictions have a root-mean-square error of 0.28 for 32 such amines. This appears to be an improvement over all other known computer-based methods. However, the method employs two empirical corrections, which would need to be expanded to accommodate other classes of ringed compounds. The 30-year-old computer-free group-additivity-based scheme by Perrin, Dempsey, and Serjeant (PDS) produced an rms error of 0.33 for the same amine set, and with updated

preference for gauche forms,35,36 likely did not achieve full equilibration, because bizarrely asymmetric distributions of

Table 4. pKa Results: Comparison of Continuum-Solvation Procedures COSMO

Jaguar

amine

expt.a

SHE

2010b

2002c

2010-1d

2010-2e

2006-1f

2006-2g

methylamine dimethylamine trimethylamine MEA morpholine

10.66 10.73 9.80 9.50 8.50

10.52 10.74 9.66 9.32 8.25

10.52 10.71 10.12

10.5 10.9 10.1 9.8 9.5

11.97 10.94 8.93

11.09 9.64 7.73

11.71 10.67 8.63 10.29 8.36

11.71 11.67 10.63 10.29 9.36

0.2

0.2

0.5

0.9

1.4

0.8

0.9

rms error

COSMO-RS

Reference 23. bReference 12. cReference 22. dReference 11. Table 3 “bare” data. eReference 11. Table 3 “H2O” data. fReference 10. Table 2 “pKa calc” data. gReference 10. Table 2 “pKa corrected” data. a

D



Article

Table 5. SHE versus PDS Predictions for pKa of 32 CO2-Relevant Amines amine

amine label

SHE0a

SHE

old PDS

new PDS

expt.

refb

NH(C2H5)2 NH2(CH2)4NH2 NH(CH3)2 N(C2H5)3 NH2C(CH3)3 NH2CH2CH3 NH2CH(CH3)2 NH2CH2CH2CH2CH3 NH2CH2CH2CH3 NH2(CH2)3NH2 NH2CH3 NH2CH2CH2CH2OH N(CH3)3 NH2(CH2)2NH2 NH(CH3)CH2CH2OH N(C2H5)2CH2CH2OH NH2C(CH3)2CH2OH NH2CH2CH2OH NH2CH2CH(CH3)OH N(CH3)2CH2CH2OH NH(CH2CH2OH)2 N(CH3)(CH2CH2OH)2 N(CH2CH2OH)3 HN(CH2CH2)2NH C4H9N2(CH3) C4H9N2(C2H5) C4H9N2(CH3) C4H9N2(C2H4OH) C4H8N2(CH3)2 HN(CH2CH2)2O C2H5N(CH2CH2)2O CH3N(CH2CH2)2O

diethylamine 1,4-butanediamine dimethylamine triethylamine tert-butylamine ethylamine iso-propylamine butylamine propylamine 1,3-propanediamine methylamine MPA trimethylamine 1,2-ethanediamine MMEA DEMEA AMP MEA MIPA DMMEA DEA MDEA TEA PPZ (piperazine) 2-MePPZ (H+ on 4) 1-EtPPZ 1-MePPZ 1-(2-EtOH)PPZ 1,4-Me2PPZ MOR (morpholine) 4-EtMOR 4-MeMOR

11.88 11.42 11.39 11.78 10.90 11.28 11.13 11.48 11.39 11.34 11.17 10.90 10.31 10.99 10.75 10.66 9.76 9.97 9.76 9.76 9.47 8.64 8.53 11.69 11.60 11.12 11.17 10.96 9.94 9.95 9.16 8.76

11.18 10.72 10.69 11.08 10.20 10.58 10.43 10.78 10.69 10.64 10.47 10.20 9.61 10.29 10.05 9.96 9.06 9.27 9.06 9.06 8.77 7.94 7.83 9.99 9.90 9.42 9.47 9.26 8.24 8.25 7.46 7.06

11.15 10.94 10.75 10.50 10.77 10.77 10.77 10.77 10.77 10.75 10.57 10.33 9.90 10.27 9.85 9.40 9.67 9.67 9.67 9.00 8.95 8.10 7.20 9.85 9.55 9.55 9.55 9.55 9.00 8.95 8.30 8.10

11.10 10.76 10.70 10.60 10.60 10.60 10.60 10.60 10.60 10.54 10.40 10.20 10.00 10.00 9.90 9.60 9.60 9.60 9.60 9.20 9.10 8.40 7.60 9.40 9.10 9.10 9.10 9.10 8.70 8.30 7.80 7.60

10.84 10.80 10.73 10.75 10.68 10.65 10.63 10.56 10.54 10.55 10.66 9.96 9.80 9.92 9.85 9.75 9.68 9.50 9.45 9.22 8.88 8.56 7.78 9.73 9.57 9.20 9.14 9.09 8.38 8.50 7.67 7.38

23 23 23 23 23 23 23 23 23 23 23 43 23 23 33 33 33 23 33 33 42 41 23 23 40 40 40 40 40 23 23 23

1.11

0.28

0.33

0.18

rms error a

b

SHE procedure without empirical corrections C in eq 3. Reference for experimental value.

issues, and hence should be the method of choice for aliphatic amines. Future work should aim at updating the values of other PDS parameters for application to other bases and to acids.

Table 6. pKa Errors in SHE versus PDS Predictions Outside the Training Set amine

pKa(expt)a

SHE error

old PDS error

new PDS error

piperidine 1-methylpyrrolidine cis-2,5-dimethylpiperazine 2-methoxyethylamine 2-(ethylamino)ethanol 2-(proylamino)ethanol 2-(butylamino)ethanol pentylamine hexylamine diethylmethylamine 3-methyl-1-butylamine

11.12 10.46 9.66 9.40 9.99 9.90 9.92 10.63 10.56 10.35 10.60

−0.49 −0.88 0.09 −0.57 −0.01 0.18 0.26 0.18 0.27 −0.05 0.27

0.23 0.04 0.19 0.17 0.06 0.15 0.13 0.14 0.21 −0.05 −0.10

−0.02 −0.06 −0.26 −0.20 0.11 0.20 0.18 −0.03 0.04 0.05 0.00

0.39

0.15

0.14

rms error a

■

ASSOCIATED CONTENT

* Supporting Information S

Images of all conformers used for data in Tables 3, 5, and 6. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS H. B. Schlegel is thanked for a stimulating 2009 Halifax postersession discussion on pKa computations. Computations were performed via Westgrid supercomputers, as well as our CFIfunded Ciaratech cluster in the Laboratory of Computational Discovery (Regina; John Jorgensen, sysadmin). An NSERC Discovery Grant (Canada) was also of assistance.

Reference 23, except reference 13 for entries 5−7.

parameter values it produced an rms error of only 0.18. The updated PDS method outperforms the best continuumsolvation methods in both speed and accuracy, can extend to other ringed amines, and has no conformer or cavity-radii E



■

Article

B.; Petersson, G. A.; Nakatsuji, H.; Caricato, M.; Li, X.; Hratchian, H. P.; Izmaylov, A. F.; Bloino, J.; Zheng, G.; Sonnenberg, J. L.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.; Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Vreven, T.; Montgomery, J. A., Jr.; Peralta, J. E.; Ogliaro, F.; Bearpark, M.; Heyd, J. J.; Brothers, E.; Kudin, K. N.; Staroverov, V. N.; Kobayashi, R.; Normand, J.; Raghavachari, K.; Rendell, A.; Burant, J. C.; Iyengar, S. S.; Tomasi, J.; Cossi, M.; Rega, N.; Millam, N. J.; Klene, M.; Knox, J. E.; Cross, J. B.; Bakken, V.; Adamo, C.; Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.; Austin, A. J.; Cammi, R.; Pomelli, C.; Ochterski, J. W.; Martin, R. L.; Morokuma, K.; Zakrzewski, V. G.; Voth, G. A.; Salvador, P.; Dannenberg, J. J.; Dapprich, S.; Daniels, A. D.; Farkas, Ö .; Foresman, J. B.; Ortiz, J. V.; Cioslowski, J.; Fox, D. J. Gaussian 09, revision B.01; Gaussian, Inc.: Wallingford, CT, 2010. (18) Perrin, D. D.; Dempsey, B.; Serjeant, E. P. pKa Prediction for Organic Acids and Bases; Chapman and Hall: New York, 1981. (19) A relevant example is: Hall, H. K. Correlation of the Base Strengths of Amines. J. Am. Chem. Soc. 1957, 79, 5441. (20) Tissandier, M. D.; Cowen, K. A.; Feng, W. Y.; Gundlach, E.; Cohen, M. H.; Earhart, A. D.; Coe, J. V.; Tuttle, T. R., Jr. The Proton’s Absolute Aqueous Enthalpy and Gibbs Free Energy of Solvation from Cluster-Ion Solvation Data. J. Phys. Chem. A 1998, 102, 7787. (21) Ben-Naim, A.; Mazo, R. M. Size Dependence of the Solvation Free Energies of Large Solutes. J. Phys. Chem. 1993, 97, 10829. (22) Klicic, J. J.; Friesner, R. A.; Liu, S.-Y.; Guida, W. C. Accurate Prediction of Acidity Constants in Aqueous Solution via Density Functional Theory and Self-Consistent Reaction Field Methods. J. Phys. Chem. A 2002, 106, 1327. (23) Haynes, W. M. CRC Handbook of Chemistry and Physics, 92nd ed.; CRC Press/Taylor and Francis: Boca Raton, FL, 2011. (24) Sadlej-Sosnowska, N. Calculation of Acidic Dissociation Constants in Water: Solvation Free Energy Terms. Their Accuracy and Impact. Theor. Chem. Acc. 2007, 118, 281. (25) Behjatmanesh-Ardakani, R.; Karimi, M. A.; Ebady, A. Cavity Shape Effect on pKa Prediction of Small Amines. J. Mol. Struct. (THEOCHEM) 2009, 910, 99. (26) Clemente, F. R., private communication. (27) Scalmani, G.; Frisch, M. Continuous Surface Charge Polarizable Continuum Models of Solvation. I. General formalism. J. Chem. Phys. 2010, 132, 114110. (28) Atom volumes were determined by a group contribution scheme, fitted to the total volumes of the 17 bases as calculated by Gaussian’s overlapping-spheres procedure. (29) Marten, B.; Kim, K.; Cortis, C.; Friesner, R. A.; Murphy, R. B.; Ringnalda, M. N.; Sitkoff, D.; Honig, B. New Model for Calculation of Solvation Free Energies: Correction of Self-Consistent Reaction Field Continuum Dielectric Theory for Short-Range Hydrogen-Bonding Effects. J. Phys. Chem. 1996, 100, 11775. (30) Head-Gordon, T. Is Water Structure around Hydrophobic Groups Clathrate-like? Proc. Natl. Acad. Sci. U.S.A. 1995, 92, 8308. (31) Hawkins, R. E.; Davidson, D. W. Dielectric Relaxation in the Clathrate Hydrates of Some Cyclic Ethers. J. Phys. Chem. 1966, 70, 1889. (32) Effects of up to 4 pKa units were reported in a continuumsolvation-based computational study of serotonin, see: Pratuangdejkul, J.; Nosoongnoen, W.; Guérin, G.-A.; Loric, S.; Conti, M.; Launay, J.M.; Manivet, P. Conformational Dependence of Serotonin Theoretical pKa Prediction. Chem. Phys. Lett. 2006, 420, 538. (33) Hamborg, E. S.; Versteeg, G. F. Dissociation Constants and Thermodynamic Properties of Amines and Alkanolamines from (293 to 353) K. J. Chem. Eng. Data 2009, 54, 1318. (34) Han, B.; Zhou, C.; Wu, J.; Tempel, D. J.; Cheng, H. Understanding CO2 Capture Mechanisms in Aqueous Monoethanolamine Via First Principles Simulations. J. Phys. Chem. Lett. 2011, 2, 522. (35) Lopez-Rendon, R.; Mora, M. A.; Alejandre, J.; Tuckerman, M. E. Molecular Dynamics Simulations of Aqueous Solutions of Ethanolamines. J. Phys. Chem. B 2006, 110, 14652.

REFERENCES

(1) Versteeg, G. F.; van Dijck, L. A. J.; van Swaaij, W. P. M. On The Kinetics Between CO2 and Alkanolamines Both in Aqueous and NonAqueous Solutions. An Overview. Chem. Eng. Commun. 1996, 144, 113. (2) da Silva, E. F.; Svendsen, H. F. Prediction of the pKa Values of Amines Using ab Initio Methods and Free-Energy Perturbations. Ind. Eng. Chem. Res. 2003, 42, 4414. (3) Alongi, K. S.; Shields, G. C. Theoretical Calculations of Acid Dissociation Constants: A Review Article. Annu. Rep. Comput. Chem. 2010, 6, 113. (4) Ho, J.; Coote, M. L. A Universal Approach for Continuum Solvent pKa Calculations: Are We There Yet? Theor. Chem. Acc. 2010, 125, 3. (5) Brown, T. N.; Mora-Diez, N. Computational Determination of Aqueous pKa Values of Protonated Benzimidazoles (Part 1). J. Phys. Chem. B 2006, 110, 9270. (6) Lu, H.; Chen, X.; Zhan, C.-G. First-Principles Calculation of pKa for Cocaine, Nicotine, Neurotransmitters, and Anilines in Aqueous Solution. J. Phys. Chem. B 2007, 111, 10599. (7) Khalili, F.; Henni, A.; East, A. L. L. Entropy Contributions In pKa Computation: Application to Alkanolamines and Piperazines. J. Mol. Struct. (THEOCHEM) 2009, 916, 1. (8) Bryantsev, V. S.; Diallo, M. S.; Goddard, W. A. pKa Calculations of Aliphatic Amines, Diamines, and Aminoamides Via Density Functional Theory with a Poisson-Boltzmann Continuum Solvent Model. J. Phys. Chem. A 2007, 111, 4422. (9) Ulander, J.; Broo, A. Use of Empirical Correction Terms in Calculating Ionization Constants. Int. J. Quantum Chem. 2005, 105, 866. (10) Eckert, F.; Klamt, A. Accurate Prediction of Basicity in Aqueous Solution with COSMO-RS. J. Comput. Chem. 2006, 27, 11. (11) Eckert, F.; Diedenhofen, M.; Klamt, A. Towards First Principles Prediction of pKa and the Cluster-Continuum Approach. Mol. Phys. 2010, 108, 229. (12) Zhang, S.; Baker, J.; Pulay, P. A Reliable and Efficient First Principles-Based Method for Predicting pKa Value. 1. Methodology. J. Phys. Chem. A 2010, 114, 425. (13) Yamada, H.; Shimizu, S.; Okabe, H.; Matsuzaki, Y.; Chowdhury, F. A.; Fujioka, Y. Prediction of the Basicity of Aqueous Amine Solutions and the Species Distribution in the Amine−H2O−CO2 System Using the COSMO-RS Method. Ind. Eng. Chem. Res. 2010, 49, 2449. (14) Seybold, P. G. Analysis of the pKas of Aliphatic Amines Using Quantum Chemical Descriptors. Int. J. Quantum Chem. 2008, 108, 2849. (15) Burger, S. K.; Liu, S.; Ayers, P. W. Practical Calculation of Molecular Acidity with the Aid of a Reference Molecule. J. Phys. Chem. A 2011, 115, 1293. (16) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Montgomery, J. A., Jr.; Vreven, T.; Kudin, K. N.; Burant, J. C.; Millam, J. M.; Iyengar, S. S.; Tomasi, J.; Barone, V.; Mennucci, B.; Cossi, M.; Scalmani, G.; Rega, N.; Petersson, G. A.; Nakatsuji, H.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.; Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Klene, M.; Li, X.; Knox, J. E.; Hratchian, H. P.; Cross, J. B.; Adamo, C.; Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.; Austin, A. J.; Cammi, R.; Pomelli, C.; Ochterski, J. W.; Ayala, P. Y.; Morokuma, K.; Voth, G. A.; Salvador, P.; Dannenberg, J. J.; Zakrzewski, V. G.; Dapprich, S.; Daniels, A. D.; Strain, M. C.; Farkas, O.; Malick, D. K.; Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Ortiz, J. V.; Cui, Q.; Baboul, A. G.; Clifford, S.; Cioslowski, J.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.; Peng, C. Y.; Nanayakkara, A.; Challacombe, M.; Gill, P. M. W.; Johnson, B.; Chen, W.; Wong, M. W.; Gonzalez, C.; Pople, J. A. Gaussian 03, revision E.01; Gaussian Inc.: Wallingford, CT, 2004. (17) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Scalmani, G.; Barone, V.; Mennucci, F



Article

(36) da Silva, E. F.; Kuznetsova, T.; Kvamme, B.; Merz, K. M., Jr. Molecular Dynamics Study of Ethanolamine as a Pure Liquid and in Aqueous Solution. J. Phys. Chem. B 2007, 111, 3695. (37) Gubskaya, A. V.; Kusalik, P. G. Molecular Dynamics Simulation Study of Ethylene Glycol, Ethylenediamine, and 2-Aminoethanol. 2. Structure in Aqueous Solutions. J. Phys. Chem. A 2004, 108, 7165. (38) We have unpublished density-functional-theory simulation results for aqueous MEA, which show that both trans and gauche MEA can persist for 8000 fs time-steps. The trans MEA produced a lower average energy, although not with statistical certainty. (39) Omura, Y.; Shimanouchi, T. Raman Spectra and Rotational Isomerism of Ethylenediammonium and Monoethanolammonium Ions in Aqueous Solutions. J. Mol. Spectrosc. 1975, 55, 430. (40) Khalili, F.; Henni, A.; East, A. L. L. pKa Values of Some Piperazines at (298, 303, 313, and 323) K. J. Chem. Eng. Data 2009, 54, 2914. (41) Hamborg, E. S.; Niederer, J. P. M.; Versteeg, G. F. Dissociation Constants and Thermodynamic Properties of Amino Acids Used in CO2 Absorption from (293 to 353) K. J. Chem. Eng. Data 2007, 52, 2491. (42) Bower, V. E.; Robinson, R. A.; Bates, R. G. Acidic Dissociation Constant and related Thermodynamic Quantities for Diethanolammonium Ion in Water from 0° to 50 °C. J. Res. Natl. Bur. Stand. 1962, 66A, 71. (43) Schwabe, K.; Graichen, W.; Spiethoff, D. Physicochemical Investigations of Alkanolamines. Z. Phys. Chem. (Munich) 1959, 20, 68.

G


Predicting pKa of Amines for CO2 Capture: Computer versus Pencil

Recommend Documents